The Agent Skills Directory

[SAFE]: The skill proactively implements security auditing for AI agent skills. It includes a normative 'Artifact Audit' section in rules/anti-patterns.md and rules/validation-and-iteration.md that requires checks for hardcoded credentials, unauthorized network access, and broad filesystem operations in any skill being processed.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a data ingestion surface as its primary function is to analyze and refactor external skill folders.
Ingestion points: The skill reads files from a user-specified directory (the target skill folder) through the provided scripts/validate-skill-maker.mjs and agentic file-system tools.
Boundary markers: Instructions explicitly state that all retrieved material must be treated as evidence and never as executable instruction authority (SKILL.md §Authority).
Capability inventory: The agent uses file-read, file-edit, and shell-search tools. The bundled validator script uses node:fs for local file access.
Sanitization: The validator performs structural checks on frontmatter, links, and code fences, and verifies source metadata dates against current time to prevent future-dated claims.
[DYNAMIC_CONTEXT_INJECTION]: The skill documents the platform-specific !command`` syntax in references/official/anthropic.md for educational purposes, describing how Claude Code handles dynamic content. The skill itself does not employ this syntax and no executable dynamic context triggers were found in the instructions.

skill-maker